Raw LLM Responses

I think this is an excellent example of how statistics and large numbers work. A single individual is extremely hard to predict, but large groups of people are relatively easy, meaning you can predict with a fair degree of accuracy the average behavior of a large group of people. We do this in business all the time to predict people's buying habits. I used to run a small business and I would use some rudimentary statistical analysis to predict how we'd do in any given year, set prices, manage inventory, etc . . . and I could generally predict to within a few percent the total profits we would have during a year based on numbers I would have at the beginning of the year. I couldn't tell you what a single customer would do, but I could predict what people would do as a whole months in advanced if I had the right information. The Twitter info doesn't surprise me in the least bit. The number of people involved means that extremes are effectively controlled for and you'll probably get a fairly average group of people, which means the common wisdom of what people may commit crimes will be what shows up in the data. If that common wisdom is at all accurate, then the data from Twitter will be fairly accurate. If that data is inaccurate, then it will be inaccurate. It isn't surprising that Twitter ended up with similar numbers as the algorithm, since the algorithm seemed to take into account things that most people would consider (prior convictions, education, socio-economic background, etc . . . ) I don't know how accurate they both were, but I'd be surprised if the Twitter poll was extremely different.